Study On The Construction Of Human Energy Consumption Model Based On Accelerometers With Different Wearing Positions

Background and Purpose: Based on multi-axis sensor wearable equipment,monitoring physical activity has become an important means.There is a lack of corresponding prediction models for monitoring physical activity with wearable equipment on different parts of the body.This study aims to construct a model for predicting energy consumption based on three-axis accelerometers worn on different parts of the body,in order to improve the accuracy and reliability of wearable devices.Methods: 40 university students were recruited,with an average age of 24.4 years old,19 females,and 21 males.30 test data were used to construct the model,and 10 test data were used to validate the model.Participants wore three-axis accelerometers on their right wrist,left wrist,chest,right hip,right ankle,and left ankle,respectively.The energy consumption during exercise was tested using the Italian COSMED Quark PFT ergo pulmonary function test device,and energy consumption data was collected.The exercise forms were walking at different speeds(2km/h,3km/h,4km/h,5km/h,6km/h)and running(7km/h,8km/h,9km/h)on the treadmill,with each speed lasting4 minutes and sufficient rest between speeds to recover to a heart rate close to resting.SPSS 21.0 statistical software was used to analyze the data and construct the model.Results:(1)The energy consumption models based on the right and left wrist-worn three-axis accelerometers were respectively: METs = 8.527 * vector sum-0.042 ×body fat rate-4.082(the adjusted R2 = 0.827,linear hypothesis test P < 0.01,regression coefficient test P < 0.01,and the equation has statistical significance),METs = 8.631 × vector sum-0.208 × BMI-0.796 × gender + 0.705(the adjusted R2 = 0.807,linear hypothesis test P < 0.01,regression coefficient test P <0.01,and the equation has statistical significance);The energy consumption model based on the chest-worn three-axis accelerometer was: METs = 27.932 * vector sum-0.092 * body fat rate-22.420(the adjusted R2 = 0.671,linear hypothesis test P < 0.01,regression coefficient test P < 0.01,and the equation has statistical significance);The energy consumption model based on the right hip-worn three-axis accelerometer was:METs = 29.275 * vector sum-0.086 * body fat rate-0.020 * body weight(kg)-22.361(the adjusted R2 = 0.752,linear hypothesis test P < 0.01,regression coefficient test P < 0.01,and the equation has statistical significance);The energy consumption models based on the right and left ankle-worn three-axis accelerometers were respectively: METs = 7.772 * vector sum + 0.233 * body fat rate-0.387 * body weight(kg)+ 0.479 * muscle mass-8.654(the adjusted R2 = 0.894,linear hypothesis test P < 0.01,regression coefficient test P < 0.01,and the equation has statistical significance),METs = 7.826 * vector sum-0.068 * body fat rate-0.196 * age +0.437(the adjusted R2 = 0.885,linear hypothesis test P < 0.01,regression coefficient test P < 0.01,and the equation has statistical significance).(2)The consistency between the predicted values and the measured values of the prediction model based on 6 human body wearing positions is good,and the consistency between the predicted values and the measured values of the joint prediction of wearing the right wrist and left ankle is also good.The accuracy of the prediction of energy consumption by accelerometers worn in different positions is as follows: for the right wrist,the MSE,RMSE,MAE,MAPE,and R2 are 1.39,1.18,0.95,16.62%,and 0.79,respectively;for the left wrist,the MSE,RMSE,MAE,MAPE,and R2 are 2.42,1.56,1.17,19.20%,and 0.63,respectively;for the chest,the MSE,RMSE,MAE,MAPE,and R2 are 2.99,1.73,1.33,22.31%,and 0.55,respectively;for the right hip,the MSE,RMSE,MAE,MAPE,and R2 are 2.09,1.45,1.17,20.28%,and 0.68,respectively;for the right ankle,the MSE,RMSE,MAE,MAPE,and R2 are 1.18,1.09,0.83,15.18%,and 0.82,respectively;for the left ankle,the MSE,RMSE,MAE,MAPE,and R2 are 1.07,1.03,0.74,12.54%,and 0.84,respectively.For the walking/running movement,the accelerometer worn at the ankle position has a high accuracy in predicting energy consumption.(3)The model for predicting energy consumption by combining the data of the right wrist and left ankle worn accelerometers is: METs = 4.968 x left ankle vector sum +3.565 x right hand vector sum-0.050 x body fat rate-4.988(the adjusted R2 of the model is 0.904,the linear hypothesis test P < 0.01,the regression coefficient test P <0.01,and the equation has statistical significance).The accuracy of this model in predicting energy consumption is: MSE,RMSE,MAE,MAPE,and R2 are 0.95,0.97,0.73,12.93%,and 0.86,respectively.Conclusion: There is a difference in the accuracy of predicting energy consumption by three-axis accelerometers worn in different human body positions,and the accuracy of predicting energy consumption by accelerometers worn at the ankle position is the highest.At the same time,the accuracy of predicting energy consumption by combining the data of the right wrist and left ankle worn accelerometers is obviously higher than that of single prediction,indicating that the joint prediction model can improve the prediction accuracy.Therefore,in the design and development of wearable equipment,it is necessary to consider the influence of different wearing positions and use appropriate models for energy consumption prediction to improve the accuracy and reliability of prediction.